lucidrains / audiolm-pytorch

Implementation of AudioLM, a SOTA Language Modeling Approach to Audio Generation out of Google Research, in Pytorch
MIT License
2.36k stars 255 forks source link

Add .eval when exporting flac #132

Closed yygle closed 1 year ago

yygle commented 1 year ago

@lucidrains Thanks for the awesome work. You may forgot to add .eval() when exporting the flac files in soundstream trainer. And it may cause differences for the output wavs.

lucidrains commented 1 year ago

@yygle oh, wasn't that done here? or are you referring to somewhere else?

yygle commented 1 year ago

sorry, my mistake. It use different names, self.soundstream and self.model.

lucidrains commented 1 year ago

@yygle yea, there is an EMA version

no worries, thanks for raising the issue!